At the AIE 2026 Shanghai conference, multiple speakers pointed to the same conclusion: the better AI performs, the more exposed companies become to ambiguous inputs, inconsistent standards, and undocumented experience. Model capability is no longer scarce; whether an organization can articulate its judgment clearly will determine whether AI can genuinely enter the business.
What this is
The conference divided AI’s limitations into three layers. The first is decision-making: a model can provide options and the basis for judgment, but it cannot bear the final decision for a company. The second is soft knowledge—experience that has not yet been documented and remains scattered across veteran employees and supply-chain relationships. The third is business definitions: for example, if “shipment” means different things to workshop, finance, and management teams, the facts AI receives will be inaccurate.
Former Cursor designer Ryo’s observation that “an answer being complete does not mean the thinking is complete” echoes Dr. Yang Ying’s view of vertical domains. General-purpose AI works like a general-purpose decoder: it can handle literature, data, and domains with clear rules. Industries such as beauty and manufacturing, however, depend on contextual experience and need to convert private knowledge into reusable methods.
Industry view
Optimists argue that AI’s greatest value is not replacing people but amplifying organizational capabilities. Once repetitive work is compressed, employees can move toward judgment, communication, and creation. Vertical industries that first organize their experience and build data assets may also gain more durable advantages.
The risks are equally clear. In the manufacturing examples, the jobs easiest to quantify are often also the ones most exposed to automation pressure. Standardizing business definitions can improve system inputs, but it may also compress the “gray zones” that once depended on manual coordination and provoke organizational resistance. The efficiency gains presented at the conference also leave unanswered whether consumer demand can grow at the same pace or how the social division of labor will adjust. Companies whose moats depend on soft knowledge must also face the risk that digitizing that experience will lower their barriers to entry.
Impact on regular people
For enterprise IT: project priorities will shift from model selection to data governance, business definition, and permission design. Companies must first align key metrics before connecting agents—AI programs that autonomously invoke tools to complete tasks. Otherwise, more advanced models will merely amplify chaos.
For individual careers: value will increasingly shift from “completing the process by hand” to “defining the problem, validating the result, and owning the judgment.” Whether experience can be articulated and reused will directly affect a professional’s moat, while simply improving operational speed will lose significance.
For consumer markets: AI will increase personalized recommendations and content supply, but whether products genuinely improve the experience will still depend on whether vendors understand the details of each use case. Consumers may receive more choices, but they will also bear the costs of filtering, verification, and privacy trade-offs.